DeepSeek released preview versions of its new flagship AI models, V4 Flash and
V4 Pro, marking major upgrades a year after disrupting Silicon Valley with its
earlier platform. The models include details on pricing, parameters, and a
maximum output of 384,000 tokens. The Hangzhou-based startup previously
triggered a trillion-dollar market selloff in 2025 with its R1 open-source
reasoning model, which rivaled top systems at far lower cost. The release had
initially raised doubts over massive AI spending, which has since rebounded,
with US tech firms projected to invest about $650 billion in AI infrastructure
in 2026.
